xref: /freebsd/sys/netinet6/nd6_rtr.c (revision 6ae1554a5d9b318f8ad53ccc39fa5a961403da73)
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	$KAME: nd6_rtr.c,v 1.111 2001/04/27 01:37:15 jinmei Exp $
30  */
31 
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
34 
35 #include "opt_inet.h"
36 #include "opt_inet6.h"
37 
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/malloc.h>
41 #include <sys/mbuf.h>
42 #include <sys/socket.h>
43 #include <sys/sockio.h>
44 #include <sys/time.h>
45 #include <sys/kernel.h>
46 #include <sys/lock.h>
47 #include <sys/errno.h>
48 #include <sys/rwlock.h>
49 #include <sys/syslog.h>
50 #include <sys/queue.h>
51 
52 #include <net/if.h>
53 #include <net/if_var.h>
54 #include <net/if_types.h>
55 #include <net/if_dl.h>
56 #include <net/route.h>
57 #include <net/radix.h>
58 #include <net/vnet.h>
59 
60 #include <netinet/in.h>
61 #include <net/if_llatbl.h>
62 #include <netinet6/in6_var.h>
63 #include <netinet6/in6_ifattach.h>
64 #include <netinet/ip6.h>
65 #include <netinet6/ip6_var.h>
66 #include <netinet6/nd6.h>
67 #include <netinet/icmp6.h>
68 #include <netinet6/scope6_var.h>
69 
70 static int rtpref(struct nd_defrouter *);
71 static struct nd_defrouter *defrtrlist_update(struct nd_defrouter *);
72 static int prelist_update(struct nd_prefixctl *, struct nd_defrouter *,
73     struct mbuf *, int);
74 static struct in6_ifaddr *in6_ifadd(struct nd_prefixctl *, int);
75 static struct nd_pfxrouter *pfxrtr_lookup(struct nd_prefix *,
76 	struct nd_defrouter *);
77 static void pfxrtr_add(struct nd_prefix *, struct nd_defrouter *);
78 static void pfxrtr_del(struct nd_pfxrouter *);
79 static struct nd_pfxrouter *find_pfxlist_reachable_router
80 (struct nd_prefix *);
81 static void defrouter_delreq(struct nd_defrouter *);
82 static void nd6_rtmsg(int, struct rtentry *);
83 
84 static int in6_init_prefix_ltimes(struct nd_prefix *);
85 static void in6_init_address_ltimes(struct nd_prefix *,
86 	struct in6_addrlifetime *);
87 
88 static int nd6_prefix_onlink(struct nd_prefix *);
89 static int nd6_prefix_offlink(struct nd_prefix *);
90 
91 static int rt6_deleteroute(struct rtentry *, void *);
92 
93 VNET_DECLARE(int, nd6_recalc_reachtm_interval);
94 #define	V_nd6_recalc_reachtm_interval	VNET(nd6_recalc_reachtm_interval)
95 
96 static VNET_DEFINE(struct ifnet *, nd6_defifp);
97 VNET_DEFINE(int, nd6_defifindex);
98 #define	V_nd6_defifp			VNET(nd6_defifp)
99 
100 VNET_DEFINE(int, ip6_use_tempaddr) = 0;
101 
102 VNET_DEFINE(int, ip6_desync_factor);
103 VNET_DEFINE(u_int32_t, ip6_temp_preferred_lifetime) = DEF_TEMP_PREFERRED_LIFETIME;
104 VNET_DEFINE(u_int32_t, ip6_temp_valid_lifetime) = DEF_TEMP_VALID_LIFETIME;
105 
106 VNET_DEFINE(int, ip6_temp_regen_advance) = TEMPADDR_REGEN_ADVANCE;
107 
108 /* RTPREF_MEDIUM has to be 0! */
109 #define RTPREF_HIGH	1
110 #define RTPREF_MEDIUM	0
111 #define RTPREF_LOW	(-1)
112 #define RTPREF_RESERVED	(-2)
113 #define RTPREF_INVALID	(-3)	/* internal */
114 
115 /*
116  * Receive Router Solicitation Message - just for routers.
117  * Router solicitation/advertisement is mostly managed by userland program
118  * (rtadvd) so here we have no function like nd6_ra_output().
119  *
120  * Based on RFC 2461
121  */
122 void
123 nd6_rs_input(struct mbuf *m, int off, int icmp6len)
124 {
125 	struct ifnet *ifp = m->m_pkthdr.rcvif;
126 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
127 	struct nd_router_solicit *nd_rs;
128 	struct in6_addr saddr6 = ip6->ip6_src;
129 	char *lladdr = NULL;
130 	int lladdrlen = 0;
131 	union nd_opts ndopts;
132 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
133 
134 	/*
135 	 * Accept RS only when V_ip6_forwarding=1 and the interface has
136 	 * no ND6_IFF_ACCEPT_RTADV.
137 	 */
138 	if (!V_ip6_forwarding || ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV)
139 		goto freeit;
140 
141 	/* Sanity checks */
142 	if (ip6->ip6_hlim != 255) {
143 		nd6log((LOG_ERR,
144 		    "nd6_rs_input: invalid hlim (%d) from %s to %s on %s\n",
145 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
146 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
147 		goto bad;
148 	}
149 
150 	/*
151 	 * Don't update the neighbor cache, if src = ::.
152 	 * This indicates that the src has no IP address assigned yet.
153 	 */
154 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
155 		goto freeit;
156 
157 #ifndef PULLDOWN_TEST
158 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
159 	nd_rs = (struct nd_router_solicit *)((caddr_t)ip6 + off);
160 #else
161 	IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len);
162 	if (nd_rs == NULL) {
163 		ICMP6STAT_INC(icp6s_tooshort);
164 		return;
165 	}
166 #endif
167 
168 	icmp6len -= sizeof(*nd_rs);
169 	nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
170 	if (nd6_options(&ndopts) < 0) {
171 		nd6log((LOG_INFO,
172 		    "nd6_rs_input: invalid ND option, ignored\n"));
173 		/* nd6_options have incremented stats */
174 		goto freeit;
175 	}
176 
177 	if (ndopts.nd_opts_src_lladdr) {
178 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
179 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
180 	}
181 
182 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
183 		nd6log((LOG_INFO,
184 		    "nd6_rs_input: lladdrlen mismatch for %s "
185 		    "(if %d, RS packet %d)\n",
186 		    ip6_sprintf(ip6bufs, &saddr6),
187 		    ifp->if_addrlen, lladdrlen - 2));
188 		goto bad;
189 	}
190 
191 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
192 
193  freeit:
194 	m_freem(m);
195 	return;
196 
197  bad:
198 	ICMP6STAT_INC(icp6s_badrs);
199 	m_freem(m);
200 }
201 
202 /*
203  * Receive Router Advertisement Message.
204  *
205  * Based on RFC 2461
206  * TODO: on-link bit on prefix information
207  * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
208  */
209 void
210 nd6_ra_input(struct mbuf *m, int off, int icmp6len)
211 {
212 	struct ifnet *ifp = m->m_pkthdr.rcvif;
213 	struct nd_ifinfo *ndi = ND_IFINFO(ifp);
214 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
215 	struct nd_router_advert *nd_ra;
216 	struct in6_addr saddr6 = ip6->ip6_src;
217 	int mcast = 0;
218 	union nd_opts ndopts;
219 	struct nd_defrouter *dr;
220 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
221 
222 	/*
223 	 * We only accept RAs only when the per-interface flag
224 	 * ND6_IFF_ACCEPT_RTADV is on the receiving interface.
225 	 */
226 	if (!(ndi->flags & ND6_IFF_ACCEPT_RTADV))
227 		goto freeit;
228 
229 	if (ip6->ip6_hlim != 255) {
230 		nd6log((LOG_ERR,
231 		    "nd6_ra_input: invalid hlim (%d) from %s to %s on %s\n",
232 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
233 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
234 		goto bad;
235 	}
236 
237 	if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
238 		nd6log((LOG_ERR,
239 		    "nd6_ra_input: src %s is not link-local\n",
240 		    ip6_sprintf(ip6bufs, &saddr6)));
241 		goto bad;
242 	}
243 
244 #ifndef PULLDOWN_TEST
245 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
246 	nd_ra = (struct nd_router_advert *)((caddr_t)ip6 + off);
247 #else
248 	IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len);
249 	if (nd_ra == NULL) {
250 		ICMP6STAT_INC(icp6s_tooshort);
251 		return;
252 	}
253 #endif
254 
255 	icmp6len -= sizeof(*nd_ra);
256 	nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
257 	if (nd6_options(&ndopts) < 0) {
258 		nd6log((LOG_INFO,
259 		    "nd6_ra_input: invalid ND option, ignored\n"));
260 		/* nd6_options have incremented stats */
261 		goto freeit;
262 	}
263 
264     {
265 	struct nd_defrouter dr0;
266 	u_int32_t advreachable = nd_ra->nd_ra_reachable;
267 
268 	/* remember if this is a multicasted advertisement */
269 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
270 		mcast = 1;
271 
272 	bzero(&dr0, sizeof(dr0));
273 	dr0.rtaddr = saddr6;
274 	dr0.flags  = nd_ra->nd_ra_flags_reserved;
275 	/*
276 	 * Effectively-disable routes from RA messages when
277 	 * ND6_IFF_NO_RADR enabled on the receiving interface or
278 	 * (ip6.forwarding == 1 && ip6.rfc6204w3 != 1).
279 	 */
280 	if (ndi->flags & ND6_IFF_NO_RADR)
281 		dr0.rtlifetime = 0;
282 	else if (V_ip6_forwarding && !V_ip6_rfc6204w3)
283 		dr0.rtlifetime = 0;
284 	else
285 		dr0.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
286 	dr0.expire = time_uptime + dr0.rtlifetime;
287 	dr0.ifp = ifp;
288 	/* unspecified or not? (RFC 2461 6.3.4) */
289 	if (advreachable) {
290 		advreachable = ntohl(advreachable);
291 		if (advreachable <= MAX_REACHABLE_TIME &&
292 		    ndi->basereachable != advreachable) {
293 			ndi->basereachable = advreachable;
294 			ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
295 			ndi->recalctm = V_nd6_recalc_reachtm_interval; /* reset */
296 		}
297 	}
298 	if (nd_ra->nd_ra_retransmit)
299 		ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
300 	if (nd_ra->nd_ra_curhoplimit) {
301 		if (ndi->chlim < nd_ra->nd_ra_curhoplimit)
302 			ndi->chlim = nd_ra->nd_ra_curhoplimit;
303 		else if (ndi->chlim != nd_ra->nd_ra_curhoplimit) {
304 			log(LOG_ERR, "RA with a lower CurHopLimit sent from "
305 			    "%s on %s (current = %d, received = %d). "
306 			    "Ignored.\n", ip6_sprintf(ip6bufs, &ip6->ip6_src),
307 			    if_name(ifp), ndi->chlim, nd_ra->nd_ra_curhoplimit);
308 		}
309 	}
310 	dr = defrtrlist_update(&dr0);
311     }
312 
313 	/*
314 	 * prefix
315 	 */
316 	if (ndopts.nd_opts_pi) {
317 		struct nd_opt_hdr *pt;
318 		struct nd_opt_prefix_info *pi = NULL;
319 		struct nd_prefixctl pr;
320 
321 		for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
322 		     pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
323 		     pt = (struct nd_opt_hdr *)((caddr_t)pt +
324 						(pt->nd_opt_len << 3))) {
325 			if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
326 				continue;
327 			pi = (struct nd_opt_prefix_info *)pt;
328 
329 			if (pi->nd_opt_pi_len != 4) {
330 				nd6log((LOG_INFO,
331 				    "nd6_ra_input: invalid option "
332 				    "len %d for prefix information option, "
333 				    "ignored\n", pi->nd_opt_pi_len));
334 				continue;
335 			}
336 
337 			if (128 < pi->nd_opt_pi_prefix_len) {
338 				nd6log((LOG_INFO,
339 				    "nd6_ra_input: invalid prefix "
340 				    "len %d for prefix information option, "
341 				    "ignored\n", pi->nd_opt_pi_prefix_len));
342 				continue;
343 			}
344 
345 			if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
346 			 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
347 				nd6log((LOG_INFO,
348 				    "nd6_ra_input: invalid prefix "
349 				    "%s, ignored\n",
350 				    ip6_sprintf(ip6bufs,
351 					&pi->nd_opt_pi_prefix)));
352 				continue;
353 			}
354 
355 			bzero(&pr, sizeof(pr));
356 			pr.ndpr_prefix.sin6_family = AF_INET6;
357 			pr.ndpr_prefix.sin6_len = sizeof(pr.ndpr_prefix);
358 			pr.ndpr_prefix.sin6_addr = pi->nd_opt_pi_prefix;
359 			pr.ndpr_ifp = (struct ifnet *)m->m_pkthdr.rcvif;
360 
361 			pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved &
362 			    ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
363 			pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved &
364 			    ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
365 			pr.ndpr_plen = pi->nd_opt_pi_prefix_len;
366 			pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time);
367 			pr.ndpr_pltime = ntohl(pi->nd_opt_pi_preferred_time);
368 			(void)prelist_update(&pr, dr, m, mcast);
369 		}
370 	}
371 
372 	/*
373 	 * MTU
374 	 */
375 	if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
376 		u_long mtu;
377 		u_long maxmtu;
378 
379 		mtu = (u_long)ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
380 
381 		/* lower bound */
382 		if (mtu < IPV6_MMTU) {
383 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu option "
384 			    "mtu=%lu sent from %s, ignoring\n",
385 			    mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src)));
386 			goto skip;
387 		}
388 
389 		/* upper bound */
390 		maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
391 		    ? ndi->maxmtu : ifp->if_mtu;
392 		if (mtu <= maxmtu) {
393 			int change = (ndi->linkmtu != mtu);
394 
395 			ndi->linkmtu = mtu;
396 			if (change) /* in6_maxmtu may change */
397 				in6_setmaxmtu();
398 		} else {
399 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu "
400 			    "mtu=%lu sent from %s; "
401 			    "exceeds maxmtu %lu, ignoring\n",
402 			    mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src), maxmtu));
403 		}
404 	}
405 
406  skip:
407 
408 	/*
409 	 * Source link layer address
410 	 */
411     {
412 	char *lladdr = NULL;
413 	int lladdrlen = 0;
414 
415 	if (ndopts.nd_opts_src_lladdr) {
416 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
417 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
418 	}
419 
420 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
421 		nd6log((LOG_INFO,
422 		    "nd6_ra_input: lladdrlen mismatch for %s "
423 		    "(if %d, RA packet %d)\n", ip6_sprintf(ip6bufs, &saddr6),
424 		    ifp->if_addrlen, lladdrlen - 2));
425 		goto bad;
426 	}
427 
428 	nd6_cache_lladdr(ifp, &saddr6, lladdr,
429 	    lladdrlen, ND_ROUTER_ADVERT, 0);
430 
431 	/*
432 	 * Installing a link-layer address might change the state of the
433 	 * router's neighbor cache, which might also affect our on-link
434 	 * detection of adveritsed prefixes.
435 	 */
436 	pfxlist_onlink_check();
437     }
438 
439  freeit:
440 	m_freem(m);
441 	return;
442 
443  bad:
444 	ICMP6STAT_INC(icp6s_badra);
445 	m_freem(m);
446 }
447 
448 /*
449  * default router list proccessing sub routines
450  */
451 
452 /* tell the change to user processes watching the routing socket. */
453 static void
454 nd6_rtmsg(int cmd, struct rtentry *rt)
455 {
456 	struct rt_addrinfo info;
457 	struct ifnet *ifp;
458 	struct ifaddr *ifa;
459 
460 	bzero((caddr_t)&info, sizeof(info));
461 	info.rti_info[RTAX_DST] = rt_key(rt);
462 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
463 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
464 	ifp = rt->rt_ifp;
465 	if (ifp != NULL) {
466 		IF_ADDR_RLOCK(ifp);
467 		ifa = TAILQ_FIRST(&ifp->if_addrhead);
468 		info.rti_info[RTAX_IFP] = ifa->ifa_addr;
469 		ifa_ref(ifa);
470 		IF_ADDR_RUNLOCK(ifp);
471 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
472 	} else
473 		ifa = NULL;
474 
475 	rt_missmsg_fib(cmd, &info, rt->rt_flags, 0, rt->rt_fibnum);
476 	if (ifa != NULL)
477 		ifa_free(ifa);
478 }
479 
480 static void
481 defrouter_addreq(struct nd_defrouter *new)
482 {
483 	struct sockaddr_in6 def, mask, gate;
484 	struct rtentry *newrt = NULL;
485 	int error;
486 
487 	bzero(&def, sizeof(def));
488 	bzero(&mask, sizeof(mask));
489 	bzero(&gate, sizeof(gate));
490 
491 	def.sin6_len = mask.sin6_len = gate.sin6_len =
492 	    sizeof(struct sockaddr_in6);
493 	def.sin6_family = gate.sin6_family = AF_INET6;
494 	gate.sin6_addr = new->rtaddr;
495 
496 	error = in6_rtrequest(RTM_ADD, (struct sockaddr *)&def,
497 	    (struct sockaddr *)&gate, (struct sockaddr *)&mask,
498 	    RTF_GATEWAY, &newrt, RT_DEFAULT_FIB);
499 	if (newrt) {
500 		nd6_rtmsg(RTM_ADD, newrt); /* tell user process */
501 		RTFREE(newrt);
502 	}
503 	if (error == 0)
504 		new->installed = 1;
505 	return;
506 }
507 
508 struct nd_defrouter *
509 defrouter_lookup(struct in6_addr *addr, struct ifnet *ifp)
510 {
511 	struct nd_defrouter *dr;
512 
513 	TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
514 		if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr))
515 			return (dr);
516 	}
517 
518 	return (NULL);		/* search failed */
519 }
520 
521 /*
522  * Remove the default route for a given router.
523  * This is just a subroutine function for defrouter_select(), and should
524  * not be called from anywhere else.
525  */
526 static void
527 defrouter_delreq(struct nd_defrouter *dr)
528 {
529 	struct sockaddr_in6 def, mask, gate;
530 	struct rtentry *oldrt = NULL;
531 
532 	bzero(&def, sizeof(def));
533 	bzero(&mask, sizeof(mask));
534 	bzero(&gate, sizeof(gate));
535 
536 	def.sin6_len = mask.sin6_len = gate.sin6_len =
537 	    sizeof(struct sockaddr_in6);
538 	def.sin6_family = gate.sin6_family = AF_INET6;
539 	gate.sin6_addr = dr->rtaddr;
540 
541 	in6_rtrequest(RTM_DELETE, (struct sockaddr *)&def,
542 	    (struct sockaddr *)&gate,
543 	    (struct sockaddr *)&mask, RTF_GATEWAY, &oldrt, RT_DEFAULT_FIB);
544 	if (oldrt) {
545 		nd6_rtmsg(RTM_DELETE, oldrt);
546 		RTFREE(oldrt);
547 	}
548 
549 	dr->installed = 0;
550 }
551 
552 /*
553  * remove all default routes from default router list
554  */
555 void
556 defrouter_reset(void)
557 {
558 	struct nd_defrouter *dr;
559 
560 	TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry)
561 		defrouter_delreq(dr);
562 
563 	/*
564 	 * XXX should we also nuke any default routers in the kernel, by
565 	 * going through them by rtalloc1()?
566 	 */
567 }
568 
569 void
570 defrtrlist_del(struct nd_defrouter *dr)
571 {
572 	struct nd_defrouter *deldr = NULL;
573 	struct nd_prefix *pr;
574 
575 	/*
576 	 * Flush all the routing table entries that use the router
577 	 * as a next hop.
578 	 */
579 	if (ND_IFINFO(dr->ifp)->flags & ND6_IFF_ACCEPT_RTADV)
580 		rt6_flush(&dr->rtaddr, dr->ifp);
581 
582 	if (dr->installed) {
583 		deldr = dr;
584 		defrouter_delreq(dr);
585 	}
586 	TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry);
587 
588 	/*
589 	 * Also delete all the pointers to the router in each prefix lists.
590 	 */
591 	LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
592 		struct nd_pfxrouter *pfxrtr;
593 		if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
594 			pfxrtr_del(pfxrtr);
595 	}
596 	pfxlist_onlink_check();
597 
598 	/*
599 	 * If the router is the primary one, choose a new one.
600 	 * Note that defrouter_select() will remove the current gateway
601 	 * from the routing table.
602 	 */
603 	if (deldr)
604 		defrouter_select();
605 
606 	free(dr, M_IP6NDP);
607 }
608 
609 /*
610  * Default Router Selection according to Section 6.3.6 of RFC 2461 and
611  * draft-ietf-ipngwg-router-selection:
612  * 1) Routers that are reachable or probably reachable should be preferred.
613  *    If we have more than one (probably) reachable router, prefer ones
614  *    with the highest router preference.
615  * 2) When no routers on the list are known to be reachable or
616  *    probably reachable, routers SHOULD be selected in a round-robin
617  *    fashion, regardless of router preference values.
618  * 3) If the Default Router List is empty, assume that all
619  *    destinations are on-link.
620  *
621  * We assume nd_defrouter is sorted by router preference value.
622  * Since the code below covers both with and without router preference cases,
623  * we do not need to classify the cases by ifdef.
624  *
625  * At this moment, we do not try to install more than one default router,
626  * even when the multipath routing is available, because we're not sure about
627  * the benefits for stub hosts comparing to the risk of making the code
628  * complicated and the possibility of introducing bugs.
629  */
630 void
631 defrouter_select(void)
632 {
633 	struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL;
634 	struct llentry *ln = NULL;
635 
636 	/*
637 	 * Let's handle easy case (3) first:
638 	 * If default router list is empty, there's nothing to be done.
639 	 */
640 	if (TAILQ_EMPTY(&V_nd_defrouter))
641 		return;
642 
643 	/*
644 	 * Search for a (probably) reachable router from the list.
645 	 * We just pick up the first reachable one (if any), assuming that
646 	 * the ordering rule of the list described in defrtrlist_update().
647 	 */
648 	TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
649 		IF_AFDATA_RLOCK(dr->ifp);
650 		if (selected_dr == NULL &&
651 		    (ln = nd6_lookup(&dr->rtaddr, 0, dr->ifp)) &&
652 		    ND6_IS_LLINFO_PROBREACH(ln)) {
653 			selected_dr = dr;
654 		}
655 		IF_AFDATA_RUNLOCK(dr->ifp);
656 		if (ln != NULL) {
657 			LLE_RUNLOCK(ln);
658 			ln = NULL;
659 		}
660 
661 		if (dr->installed && installed_dr == NULL)
662 			installed_dr = dr;
663 		else if (dr->installed && installed_dr) {
664 			/* this should not happen.  warn for diagnosis. */
665 			log(LOG_ERR, "defrouter_select: more than one router"
666 			    " is installed\n");
667 		}
668 	}
669 	/*
670 	 * If none of the default routers was found to be reachable,
671 	 * round-robin the list regardless of preference.
672 	 * Otherwise, if we have an installed router, check if the selected
673 	 * (reachable) router should really be preferred to the installed one.
674 	 * We only prefer the new router when the old one is not reachable
675 	 * or when the new one has a really higher preference value.
676 	 */
677 	if (selected_dr == NULL) {
678 		if (installed_dr == NULL || !TAILQ_NEXT(installed_dr, dr_entry))
679 			selected_dr = TAILQ_FIRST(&V_nd_defrouter);
680 		else
681 			selected_dr = TAILQ_NEXT(installed_dr, dr_entry);
682 	} else if (installed_dr) {
683 		IF_AFDATA_RLOCK(installed_dr->ifp);
684 		if ((ln = nd6_lookup(&installed_dr->rtaddr, 0, installed_dr->ifp)) &&
685 		    ND6_IS_LLINFO_PROBREACH(ln) &&
686 		    rtpref(selected_dr) <= rtpref(installed_dr)) {
687 			selected_dr = installed_dr;
688 		}
689 		IF_AFDATA_RUNLOCK(installed_dr->ifp);
690 		if (ln != NULL)
691 			LLE_RUNLOCK(ln);
692 	}
693 
694 	/*
695 	 * If the selected router is different than the installed one,
696 	 * remove the installed router and install the selected one.
697 	 * Note that the selected router is never NULL here.
698 	 */
699 	if (installed_dr != selected_dr) {
700 		if (installed_dr)
701 			defrouter_delreq(installed_dr);
702 		defrouter_addreq(selected_dr);
703 	}
704 
705 	return;
706 }
707 
708 /*
709  * for default router selection
710  * regards router-preference field as a 2-bit signed integer
711  */
712 static int
713 rtpref(struct nd_defrouter *dr)
714 {
715 	switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) {
716 	case ND_RA_FLAG_RTPREF_HIGH:
717 		return (RTPREF_HIGH);
718 	case ND_RA_FLAG_RTPREF_MEDIUM:
719 	case ND_RA_FLAG_RTPREF_RSV:
720 		return (RTPREF_MEDIUM);
721 	case ND_RA_FLAG_RTPREF_LOW:
722 		return (RTPREF_LOW);
723 	default:
724 		/*
725 		 * This case should never happen.  If it did, it would mean a
726 		 * serious bug of kernel internal.  We thus always bark here.
727 		 * Or, can we even panic?
728 		 */
729 		log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->flags);
730 		return (RTPREF_INVALID);
731 	}
732 	/* NOTREACHED */
733 }
734 
735 static struct nd_defrouter *
736 defrtrlist_update(struct nd_defrouter *new)
737 {
738 	struct nd_defrouter *dr, *n;
739 
740 	if ((dr = defrouter_lookup(&new->rtaddr, new->ifp)) != NULL) {
741 		/* entry exists */
742 		if (new->rtlifetime == 0) {
743 			defrtrlist_del(dr);
744 			dr = NULL;
745 		} else {
746 			int oldpref = rtpref(dr);
747 
748 			/* override */
749 			dr->flags = new->flags; /* xxx flag check */
750 			dr->rtlifetime = new->rtlifetime;
751 			dr->expire = new->expire;
752 
753 			/*
754 			 * If the preference does not change, there's no need
755 			 * to sort the entries. Also make sure the selected
756 			 * router is still installed in the kernel.
757 			 */
758 			if (dr->installed && rtpref(new) == oldpref)
759 				return (dr);
760 
761 			/*
762 			 * preferred router may be changed, so relocate
763 			 * this router.
764 			 * XXX: calling TAILQ_REMOVE directly is a bad manner.
765 			 * However, since defrtrlist_del() has many side
766 			 * effects, we intentionally do so here.
767 			 * defrouter_select() below will handle routing
768 			 * changes later.
769 			 */
770 			TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry);
771 			n = dr;
772 			goto insert;
773 		}
774 		return (dr);
775 	}
776 
777 	/* entry does not exist */
778 	if (new->rtlifetime == 0)
779 		return (NULL);
780 
781 	n = (struct nd_defrouter *)malloc(sizeof(*n), M_IP6NDP, M_NOWAIT);
782 	if (n == NULL)
783 		return (NULL);
784 	bzero(n, sizeof(*n));
785 	*n = *new;
786 
787 insert:
788 	/*
789 	 * Insert the new router in the Default Router List;
790 	 * The Default Router List should be in the descending order
791 	 * of router-preferece.  Routers with the same preference are
792 	 * sorted in the arriving time order.
793 	 */
794 
795 	/* insert at the end of the group */
796 	TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
797 		if (rtpref(n) > rtpref(dr))
798 			break;
799 	}
800 	if (dr)
801 		TAILQ_INSERT_BEFORE(dr, n, dr_entry);
802 	else
803 		TAILQ_INSERT_TAIL(&V_nd_defrouter, n, dr_entry);
804 
805 	defrouter_select();
806 
807 	return (n);
808 }
809 
810 static struct nd_pfxrouter *
811 pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr)
812 {
813 	struct nd_pfxrouter *search;
814 
815 	LIST_FOREACH(search, &pr->ndpr_advrtrs, pfr_entry) {
816 		if (search->router == dr)
817 			break;
818 	}
819 
820 	return (search);
821 }
822 
823 static void
824 pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr)
825 {
826 	struct nd_pfxrouter *new;
827 
828 	new = (struct nd_pfxrouter *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
829 	if (new == NULL)
830 		return;
831 	bzero(new, sizeof(*new));
832 	new->router = dr;
833 
834 	LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry);
835 
836 	pfxlist_onlink_check();
837 }
838 
839 static void
840 pfxrtr_del(struct nd_pfxrouter *pfr)
841 {
842 	LIST_REMOVE(pfr, pfr_entry);
843 	free(pfr, M_IP6NDP);
844 }
845 
846 struct nd_prefix *
847 nd6_prefix_lookup(struct nd_prefixctl *key)
848 {
849 	struct nd_prefix *search;
850 
851 	LIST_FOREACH(search, &V_nd_prefix, ndpr_entry) {
852 		if (key->ndpr_ifp == search->ndpr_ifp &&
853 		    key->ndpr_plen == search->ndpr_plen &&
854 		    in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr,
855 		    &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) {
856 			break;
857 		}
858 	}
859 
860 	return (search);
861 }
862 
863 int
864 nd6_prelist_add(struct nd_prefixctl *pr, struct nd_defrouter *dr,
865     struct nd_prefix **newp)
866 {
867 	struct nd_prefix *new = NULL;
868 	int error = 0;
869 	char ip6buf[INET6_ADDRSTRLEN];
870 
871 	new = (struct nd_prefix *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
872 	if (new == NULL)
873 		return(ENOMEM);
874 	bzero(new, sizeof(*new));
875 	new->ndpr_ifp = pr->ndpr_ifp;
876 	new->ndpr_prefix = pr->ndpr_prefix;
877 	new->ndpr_plen = pr->ndpr_plen;
878 	new->ndpr_vltime = pr->ndpr_vltime;
879 	new->ndpr_pltime = pr->ndpr_pltime;
880 	new->ndpr_flags = pr->ndpr_flags;
881 	if ((error = in6_init_prefix_ltimes(new)) != 0) {
882 		free(new, M_IP6NDP);
883 		return(error);
884 	}
885 	new->ndpr_lastupdate = time_uptime;
886 	if (newp != NULL)
887 		*newp = new;
888 
889 	/* initialization */
890 	LIST_INIT(&new->ndpr_advrtrs);
891 	in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen);
892 	/* make prefix in the canonical form */
893 	IN6_MASK_ADDR(&new->ndpr_prefix.sin6_addr, &new->ndpr_mask);
894 
895 	/* link ndpr_entry to nd_prefix list */
896 	LIST_INSERT_HEAD(&V_nd_prefix, new, ndpr_entry);
897 
898 	/* ND_OPT_PI_FLAG_ONLINK processing */
899 	if (new->ndpr_raf_onlink) {
900 		int e;
901 
902 		if ((e = nd6_prefix_onlink(new)) != 0) {
903 			nd6log((LOG_ERR, "nd6_prelist_add: failed to make "
904 			    "the prefix %s/%d on-link on %s (errno=%d)\n",
905 			    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
906 			    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
907 			/* proceed anyway. XXX: is it correct? */
908 		}
909 	}
910 
911 	if (dr)
912 		pfxrtr_add(new, dr);
913 
914 	return 0;
915 }
916 
917 void
918 prelist_remove(struct nd_prefix *pr)
919 {
920 	struct nd_pfxrouter *pfr, *next;
921 	int e;
922 	char ip6buf[INET6_ADDRSTRLEN];
923 
924 	/* make sure to invalidate the prefix until it is really freed. */
925 	pr->ndpr_vltime = 0;
926 	pr->ndpr_pltime = 0;
927 
928 	/*
929 	 * Though these flags are now meaningless, we'd rather keep the value
930 	 * of pr->ndpr_raf_onlink and pr->ndpr_raf_auto not to confuse users
931 	 * when executing "ndp -p".
932 	 */
933 
934 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 &&
935 	    (e = nd6_prefix_offlink(pr)) != 0) {
936 		nd6log((LOG_ERR, "prelist_remove: failed to make %s/%d offlink "
937 		    "on %s, errno=%d\n",
938 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
939 		    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
940 		/* what should we do? */
941 	}
942 
943 	if (pr->ndpr_refcnt > 0)
944 		return;		/* notice here? */
945 
946 	/* unlink ndpr_entry from nd_prefix list */
947 	LIST_REMOVE(pr, ndpr_entry);
948 
949 	/* free list of routers that adversed the prefix */
950 	LIST_FOREACH_SAFE(pfr, &pr->ndpr_advrtrs, pfr_entry, next) {
951 		free(pfr, M_IP6NDP);
952 	}
953 	free(pr, M_IP6NDP);
954 
955 	pfxlist_onlink_check();
956 }
957 
958 /*
959  * dr - may be NULL
960  */
961 
962 static int
963 prelist_update(struct nd_prefixctl *new, struct nd_defrouter *dr,
964     struct mbuf *m, int mcast)
965 {
966 	struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL;
967 	struct ifaddr *ifa;
968 	struct ifnet *ifp = new->ndpr_ifp;
969 	struct nd_prefix *pr;
970 	int error = 0;
971 	int newprefix = 0;
972 	int auth;
973 	struct in6_addrlifetime lt6_tmp;
974 	char ip6buf[INET6_ADDRSTRLEN];
975 
976 	auth = 0;
977 	if (m) {
978 		/*
979 		 * Authenticity for NA consists authentication for
980 		 * both IP header and IP datagrams, doesn't it ?
981 		 */
982 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
983 		auth = ((m->m_flags & M_AUTHIPHDR) &&
984 		    (m->m_flags & M_AUTHIPDGM));
985 #endif
986 	}
987 
988 	if ((pr = nd6_prefix_lookup(new)) != NULL) {
989 		/*
990 		 * nd6_prefix_lookup() ensures that pr and new have the same
991 		 * prefix on a same interface.
992 		 */
993 
994 		/*
995 		 * Update prefix information.  Note that the on-link (L) bit
996 		 * and the autonomous (A) bit should NOT be changed from 1
997 		 * to 0.
998 		 */
999 		if (new->ndpr_raf_onlink == 1)
1000 			pr->ndpr_raf_onlink = 1;
1001 		if (new->ndpr_raf_auto == 1)
1002 			pr->ndpr_raf_auto = 1;
1003 		if (new->ndpr_raf_onlink) {
1004 			pr->ndpr_vltime = new->ndpr_vltime;
1005 			pr->ndpr_pltime = new->ndpr_pltime;
1006 			(void)in6_init_prefix_ltimes(pr); /* XXX error case? */
1007 			pr->ndpr_lastupdate = time_uptime;
1008 		}
1009 
1010 		if (new->ndpr_raf_onlink &&
1011 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1012 			int e;
1013 
1014 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1015 				nd6log((LOG_ERR,
1016 				    "prelist_update: failed to make "
1017 				    "the prefix %s/%d on-link on %s "
1018 				    "(errno=%d)\n",
1019 				    ip6_sprintf(ip6buf,
1020 					    &pr->ndpr_prefix.sin6_addr),
1021 				    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
1022 				/* proceed anyway. XXX: is it correct? */
1023 			}
1024 		}
1025 
1026 		if (dr && pfxrtr_lookup(pr, dr) == NULL)
1027 			pfxrtr_add(pr, dr);
1028 	} else {
1029 		struct nd_prefix *newpr = NULL;
1030 
1031 		newprefix = 1;
1032 
1033 		if (new->ndpr_vltime == 0)
1034 			goto end;
1035 		if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0)
1036 			goto end;
1037 
1038 		error = nd6_prelist_add(new, dr, &newpr);
1039 		if (error != 0 || newpr == NULL) {
1040 			nd6log((LOG_NOTICE, "prelist_update: "
1041 			    "nd6_prelist_add failed for %s/%d on %s "
1042 			    "errno=%d, returnpr=%p\n",
1043 			    ip6_sprintf(ip6buf, &new->ndpr_prefix.sin6_addr),
1044 			    new->ndpr_plen, if_name(new->ndpr_ifp),
1045 			    error, newpr));
1046 			goto end; /* we should just give up in this case. */
1047 		}
1048 
1049 		/*
1050 		 * XXX: from the ND point of view, we can ignore a prefix
1051 		 * with the on-link bit being zero.  However, we need a
1052 		 * prefix structure for references from autoconfigured
1053 		 * addresses.  Thus, we explicitly make sure that the prefix
1054 		 * itself expires now.
1055 		 */
1056 		if (newpr->ndpr_raf_onlink == 0) {
1057 			newpr->ndpr_vltime = 0;
1058 			newpr->ndpr_pltime = 0;
1059 			in6_init_prefix_ltimes(newpr);
1060 		}
1061 
1062 		pr = newpr;
1063 	}
1064 
1065 	/*
1066 	 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
1067 	 * Note that pr must be non NULL at this point.
1068 	 */
1069 
1070 	/* 5.5.3 (a). Ignore the prefix without the A bit set. */
1071 	if (!new->ndpr_raf_auto)
1072 		goto end;
1073 
1074 	/*
1075 	 * 5.5.3 (b). the link-local prefix should have been ignored in
1076 	 * nd6_ra_input.
1077 	 */
1078 
1079 	/* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
1080 	if (new->ndpr_pltime > new->ndpr_vltime) {
1081 		error = EINVAL;	/* XXX: won't be used */
1082 		goto end;
1083 	}
1084 
1085 	/*
1086 	 * 5.5.3 (d).  If the prefix advertised is not equal to the prefix of
1087 	 * an address configured by stateless autoconfiguration already in the
1088 	 * list of addresses associated with the interface, and the Valid
1089 	 * Lifetime is not 0, form an address.  We first check if we have
1090 	 * a matching prefix.
1091 	 * Note: we apply a clarification in rfc2462bis-02 here.  We only
1092 	 * consider autoconfigured addresses while RFC2462 simply said
1093 	 * "address".
1094 	 */
1095 	IF_ADDR_RLOCK(ifp);
1096 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1097 		struct in6_ifaddr *ifa6;
1098 		u_int32_t remaininglifetime;
1099 
1100 		if (ifa->ifa_addr->sa_family != AF_INET6)
1101 			continue;
1102 
1103 		ifa6 = (struct in6_ifaddr *)ifa;
1104 
1105 		/*
1106 		 * We only consider autoconfigured addresses as per rfc2462bis.
1107 		 */
1108 		if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF))
1109 			continue;
1110 
1111 		/*
1112 		 * Spec is not clear here, but I believe we should concentrate
1113 		 * on unicast (i.e. not anycast) addresses.
1114 		 * XXX: other ia6_flags? detached or duplicated?
1115 		 */
1116 		if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
1117 			continue;
1118 
1119 		/*
1120 		 * Ignore the address if it is not associated with a prefix
1121 		 * or is associated with a prefix that is different from this
1122 		 * one.  (pr is never NULL here)
1123 		 */
1124 		if (ifa6->ia6_ndpr != pr)
1125 			continue;
1126 
1127 		if (ia6_match == NULL) /* remember the first one */
1128 			ia6_match = ifa6;
1129 
1130 		/*
1131 		 * An already autoconfigured address matched.  Now that we
1132 		 * are sure there is at least one matched address, we can
1133 		 * proceed to 5.5.3. (e): update the lifetimes according to the
1134 		 * "two hours" rule and the privacy extension.
1135 		 * We apply some clarifications in rfc2462bis:
1136 		 * - use remaininglifetime instead of storedlifetime as a
1137 		 *   variable name
1138 		 * - remove the dead code in the "two-hour" rule
1139 		 */
1140 #define TWOHOUR		(120*60)
1141 		lt6_tmp = ifa6->ia6_lifetime;
1142 
1143 		if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
1144 			remaininglifetime = ND6_INFINITE_LIFETIME;
1145 		else if (time_uptime - ifa6->ia6_updatetime >
1146 			 lt6_tmp.ia6t_vltime) {
1147 			/*
1148 			 * The case of "invalid" address.  We should usually
1149 			 * not see this case.
1150 			 */
1151 			remaininglifetime = 0;
1152 		} else
1153 			remaininglifetime = lt6_tmp.ia6t_vltime -
1154 			    (time_uptime - ifa6->ia6_updatetime);
1155 
1156 		/* when not updating, keep the current stored lifetime. */
1157 		lt6_tmp.ia6t_vltime = remaininglifetime;
1158 
1159 		if (TWOHOUR < new->ndpr_vltime ||
1160 		    remaininglifetime < new->ndpr_vltime) {
1161 			lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1162 		} else if (remaininglifetime <= TWOHOUR) {
1163 			if (auth) {
1164 				lt6_tmp.ia6t_vltime = new->ndpr_vltime;
1165 			}
1166 		} else {
1167 			/*
1168 			 * new->ndpr_vltime <= TWOHOUR &&
1169 			 * TWOHOUR < remaininglifetime
1170 			 */
1171 			lt6_tmp.ia6t_vltime = TWOHOUR;
1172 		}
1173 
1174 		/* The 2 hour rule is not imposed for preferred lifetime. */
1175 		lt6_tmp.ia6t_pltime = new->ndpr_pltime;
1176 
1177 		in6_init_address_ltimes(pr, &lt6_tmp);
1178 
1179 		/*
1180 		 * We need to treat lifetimes for temporary addresses
1181 		 * differently, according to
1182 		 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
1183 		 * we only update the lifetimes when they are in the maximum
1184 		 * intervals.
1185 		 */
1186 		if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
1187 			u_int32_t maxvltime, maxpltime;
1188 
1189 			if (V_ip6_temp_valid_lifetime >
1190 			    (u_int32_t)((time_uptime - ifa6->ia6_createtime) +
1191 			    V_ip6_desync_factor)) {
1192 				maxvltime = V_ip6_temp_valid_lifetime -
1193 				    (time_uptime - ifa6->ia6_createtime) -
1194 				    V_ip6_desync_factor;
1195 			} else
1196 				maxvltime = 0;
1197 			if (V_ip6_temp_preferred_lifetime >
1198 			    (u_int32_t)((time_uptime - ifa6->ia6_createtime) +
1199 			    V_ip6_desync_factor)) {
1200 				maxpltime = V_ip6_temp_preferred_lifetime -
1201 				    (time_uptime - ifa6->ia6_createtime) -
1202 				    V_ip6_desync_factor;
1203 			} else
1204 				maxpltime = 0;
1205 
1206 			if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
1207 			    lt6_tmp.ia6t_vltime > maxvltime) {
1208 				lt6_tmp.ia6t_vltime = maxvltime;
1209 			}
1210 			if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
1211 			    lt6_tmp.ia6t_pltime > maxpltime) {
1212 				lt6_tmp.ia6t_pltime = maxpltime;
1213 			}
1214 		}
1215 		ifa6->ia6_lifetime = lt6_tmp;
1216 		ifa6->ia6_updatetime = time_uptime;
1217 	}
1218 	IF_ADDR_RUNLOCK(ifp);
1219 	if (ia6_match == NULL && new->ndpr_vltime) {
1220 		int ifidlen;
1221 
1222 		/*
1223 		 * 5.5.3 (d) (continued)
1224 		 * No address matched and the valid lifetime is non-zero.
1225 		 * Create a new address.
1226 		 */
1227 
1228 		/*
1229 		 * Prefix Length check:
1230 		 * If the sum of the prefix length and interface identifier
1231 		 * length does not equal 128 bits, the Prefix Information
1232 		 * option MUST be ignored.  The length of the interface
1233 		 * identifier is defined in a separate link-type specific
1234 		 * document.
1235 		 */
1236 		ifidlen = in6_if2idlen(ifp);
1237 		if (ifidlen < 0) {
1238 			/* this should not happen, so we always log it. */
1239 			log(LOG_ERR, "prelist_update: IFID undefined (%s)\n",
1240 			    if_name(ifp));
1241 			goto end;
1242 		}
1243 		if (ifidlen + pr->ndpr_plen != 128) {
1244 			nd6log((LOG_INFO,
1245 			    "prelist_update: invalid prefixlen "
1246 			    "%d for %s, ignored\n",
1247 			    pr->ndpr_plen, if_name(ifp)));
1248 			goto end;
1249 		}
1250 
1251 		if ((ia6 = in6_ifadd(new, mcast)) != NULL) {
1252 			/*
1253 			 * note that we should use pr (not new) for reference.
1254 			 */
1255 			pr->ndpr_refcnt++;
1256 			ia6->ia6_ndpr = pr;
1257 
1258 			/*
1259 			 * RFC 3041 3.3 (2).
1260 			 * When a new public address is created as described
1261 			 * in RFC2462, also create a new temporary address.
1262 			 *
1263 			 * RFC 3041 3.5.
1264 			 * When an interface connects to a new link, a new
1265 			 * randomized interface identifier should be generated
1266 			 * immediately together with a new set of temporary
1267 			 * addresses.  Thus, we specifiy 1 as the 2nd arg of
1268 			 * in6_tmpifadd().
1269 			 */
1270 			if (V_ip6_use_tempaddr) {
1271 				int e;
1272 				if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
1273 					nd6log((LOG_NOTICE, "prelist_update: "
1274 					    "failed to create a temporary "
1275 					    "address, errno=%d\n",
1276 					    e));
1277 				}
1278 			}
1279 			ifa_free(&ia6->ia_ifa);
1280 
1281 			/*
1282 			 * A newly added address might affect the status
1283 			 * of other addresses, so we check and update it.
1284 			 * XXX: what if address duplication happens?
1285 			 */
1286 			pfxlist_onlink_check();
1287 		} else {
1288 			/* just set an error. do not bark here. */
1289 			error = EADDRNOTAVAIL; /* XXX: might be unused. */
1290 		}
1291 	}
1292 
1293  end:
1294 	return error;
1295 }
1296 
1297 /*
1298  * A supplement function used in the on-link detection below;
1299  * detect if a given prefix has a (probably) reachable advertising router.
1300  * XXX: lengthy function name...
1301  */
1302 static struct nd_pfxrouter *
1303 find_pfxlist_reachable_router(struct nd_prefix *pr)
1304 {
1305 	struct nd_pfxrouter *pfxrtr;
1306 	struct llentry *ln;
1307 	int canreach;
1308 
1309 	LIST_FOREACH(pfxrtr, &pr->ndpr_advrtrs, pfr_entry) {
1310 		IF_AFDATA_RLOCK(pfxrtr->router->ifp);
1311 		ln = nd6_lookup(&pfxrtr->router->rtaddr, 0, pfxrtr->router->ifp);
1312 		IF_AFDATA_RUNLOCK(pfxrtr->router->ifp);
1313 		if (ln == NULL)
1314 			continue;
1315 		canreach = ND6_IS_LLINFO_PROBREACH(ln);
1316 		LLE_RUNLOCK(ln);
1317 		if (canreach)
1318 			break;
1319 	}
1320 	return (pfxrtr);
1321 }
1322 
1323 /*
1324  * Check if each prefix in the prefix list has at least one available router
1325  * that advertised the prefix (a router is "available" if its neighbor cache
1326  * entry is reachable or probably reachable).
1327  * If the check fails, the prefix may be off-link, because, for example,
1328  * we have moved from the network but the lifetime of the prefix has not
1329  * expired yet.  So we should not use the prefix if there is another prefix
1330  * that has an available router.
1331  * But, if there is no prefix that has an available router, we still regards
1332  * all the prefixes as on-link.  This is because we can't tell if all the
1333  * routers are simply dead or if we really moved from the network and there
1334  * is no router around us.
1335  */
1336 void
1337 pfxlist_onlink_check()
1338 {
1339 	struct nd_prefix *pr;
1340 	struct in6_ifaddr *ifa;
1341 	struct nd_defrouter *dr;
1342 	struct nd_pfxrouter *pfxrtr = NULL;
1343 
1344 	/*
1345 	 * Check if there is a prefix that has a reachable advertising
1346 	 * router.
1347 	 */
1348 	LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1349 		if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
1350 			break;
1351 	}
1352 
1353 	/*
1354 	 * If we have no such prefix, check whether we still have a router
1355 	 * that does not advertise any prefixes.
1356 	 */
1357 	if (pr == NULL) {
1358 		TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
1359 			struct nd_prefix *pr0;
1360 
1361 			LIST_FOREACH(pr0, &V_nd_prefix, ndpr_entry) {
1362 				if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
1363 					break;
1364 			}
1365 			if (pfxrtr != NULL)
1366 				break;
1367 		}
1368 	}
1369 	if (pr != NULL || (!TAILQ_EMPTY(&V_nd_defrouter) && pfxrtr == NULL)) {
1370 		/*
1371 		 * There is at least one prefix that has a reachable router,
1372 		 * or at least a router which probably does not advertise
1373 		 * any prefixes.  The latter would be the case when we move
1374 		 * to a new link where we have a router that does not provide
1375 		 * prefixes and we configure an address by hand.
1376 		 * Detach prefixes which have no reachable advertising
1377 		 * router, and attach other prefixes.
1378 		 */
1379 		LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1380 			/* XXX: a link-local prefix should never be detached */
1381 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1382 				continue;
1383 
1384 			/*
1385 			 * we aren't interested in prefixes without the L bit
1386 			 * set.
1387 			 */
1388 			if (pr->ndpr_raf_onlink == 0)
1389 				continue;
1390 
1391 			if (pr->ndpr_raf_auto == 0)
1392 				continue;
1393 
1394 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1395 			    find_pfxlist_reachable_router(pr) == NULL)
1396 				pr->ndpr_stateflags |= NDPRF_DETACHED;
1397 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1398 			    find_pfxlist_reachable_router(pr) != 0)
1399 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1400 		}
1401 	} else {
1402 		/* there is no prefix that has a reachable router */
1403 		LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1404 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1405 				continue;
1406 
1407 			if (pr->ndpr_raf_onlink == 0)
1408 				continue;
1409 
1410 			if (pr->ndpr_raf_auto == 0)
1411 				continue;
1412 
1413 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1414 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
1415 		}
1416 	}
1417 
1418 	/*
1419 	 * Remove each interface route associated with a (just) detached
1420 	 * prefix, and reinstall the interface route for a (just) attached
1421 	 * prefix.  Note that all attempt of reinstallation does not
1422 	 * necessarily success, when a same prefix is shared among multiple
1423 	 * interfaces.  Such cases will be handled in nd6_prefix_onlink,
1424 	 * so we don't have to care about them.
1425 	 */
1426 	LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
1427 		int e;
1428 		char ip6buf[INET6_ADDRSTRLEN];
1429 
1430 		if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
1431 			continue;
1432 
1433 		if (pr->ndpr_raf_onlink == 0)
1434 			continue;
1435 
1436 		if (pr->ndpr_raf_auto == 0)
1437 			continue;
1438 
1439 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
1440 		    (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1441 			if ((e = nd6_prefix_offlink(pr)) != 0) {
1442 				nd6log((LOG_ERR,
1443 				    "pfxlist_onlink_check: failed to "
1444 				    "make %s/%d offlink, errno=%d\n",
1445 				    ip6_sprintf(ip6buf,
1446 					    &pr->ndpr_prefix.sin6_addr),
1447 					    pr->ndpr_plen, e));
1448 			}
1449 		}
1450 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
1451 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 &&
1452 		    pr->ndpr_raf_onlink) {
1453 			if ((e = nd6_prefix_onlink(pr)) != 0) {
1454 				nd6log((LOG_ERR,
1455 				    "pfxlist_onlink_check: failed to "
1456 				    "make %s/%d onlink, errno=%d\n",
1457 				    ip6_sprintf(ip6buf,
1458 					    &pr->ndpr_prefix.sin6_addr),
1459 					    pr->ndpr_plen, e));
1460 			}
1461 		}
1462 	}
1463 
1464 	/*
1465 	 * Changes on the prefix status might affect address status as well.
1466 	 * Make sure that all addresses derived from an attached prefix are
1467 	 * attached, and that all addresses derived from a detached prefix are
1468 	 * detached.  Note, however, that a manually configured address should
1469 	 * always be attached.
1470 	 * The precise detection logic is same as the one for prefixes.
1471 	 *
1472 	 * XXXRW: in6_ifaddrhead locking.
1473 	 */
1474 	TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1475 		if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
1476 			continue;
1477 
1478 		if (ifa->ia6_ndpr == NULL) {
1479 			/*
1480 			 * This can happen when we first configure the address
1481 			 * (i.e. the address exists, but the prefix does not).
1482 			 * XXX: complicated relationships...
1483 			 */
1484 			continue;
1485 		}
1486 
1487 		if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
1488 			break;
1489 	}
1490 	if (ifa) {
1491 		TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1492 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1493 				continue;
1494 
1495 			if (ifa->ia6_ndpr == NULL) /* XXX: see above. */
1496 				continue;
1497 
1498 			if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) {
1499 				if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1500 					ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1501 					ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1502 					nd6_dad_start((struct ifaddr *)ifa, 0);
1503 				}
1504 			} else {
1505 				ifa->ia6_flags |= IN6_IFF_DETACHED;
1506 			}
1507 		}
1508 	}
1509 	else {
1510 		TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
1511 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
1512 				continue;
1513 
1514 			if (ifa->ia6_flags & IN6_IFF_DETACHED) {
1515 				ifa->ia6_flags &= ~IN6_IFF_DETACHED;
1516 				ifa->ia6_flags |= IN6_IFF_TENTATIVE;
1517 				/* Do we need a delay in this case? */
1518 				nd6_dad_start((struct ifaddr *)ifa, 0);
1519 			}
1520 		}
1521 	}
1522 }
1523 
1524 static int
1525 nd6_prefix_onlink_rtrequest(struct nd_prefix *pr, struct ifaddr *ifa)
1526 {
1527 	static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1528 	struct radix_node_head *rnh;
1529 	struct rtentry *rt;
1530 	struct sockaddr_in6 mask6;
1531 	u_long rtflags;
1532 	int error, a_failure, fibnum;
1533 
1534 	/*
1535 	 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
1536 	 * ifa->ifa_rtrequest = nd6_rtrequest;
1537 	 */
1538 	bzero(&mask6, sizeof(mask6));
1539 	mask6.sin6_len = sizeof(mask6);
1540 	mask6.sin6_addr = pr->ndpr_mask;
1541 	rtflags = (ifa->ifa_flags & ~IFA_RTSELF) | RTF_UP;
1542 
1543 	a_failure = 0;
1544 	for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
1545 
1546 		rt = NULL;
1547 		error = in6_rtrequest(RTM_ADD,
1548 		    (struct sockaddr *)&pr->ndpr_prefix, ifa->ifa_addr,
1549 		    (struct sockaddr *)&mask6, rtflags, &rt, fibnum);
1550 		if (error == 0) {
1551 			KASSERT(rt != NULL, ("%s: in6_rtrequest return no "
1552 			    "error(%d) but rt is NULL, pr=%p, ifa=%p", __func__,
1553 			    error, pr, ifa));
1554 
1555 			rnh = rt_tables_get_rnh(rt->rt_fibnum, AF_INET6);
1556 			/* XXX what if rhn == NULL? */
1557 			RADIX_NODE_HEAD_LOCK(rnh);
1558 			RT_LOCK(rt);
1559 			if (rt_setgate(rt, rt_key(rt),
1560 			    (struct sockaddr *)&null_sdl) == 0) {
1561 				struct sockaddr_dl *dl;
1562 
1563 				dl = (struct sockaddr_dl *)rt->rt_gateway;
1564 				dl->sdl_type = rt->rt_ifp->if_type;
1565 				dl->sdl_index = rt->rt_ifp->if_index;
1566 			}
1567 			RADIX_NODE_HEAD_UNLOCK(rnh);
1568 			nd6_rtmsg(RTM_ADD, rt);
1569 			RT_UNLOCK(rt);
1570 			pr->ndpr_stateflags |= NDPRF_ONLINK;
1571 		} else {
1572 			char ip6buf[INET6_ADDRSTRLEN];
1573 			char ip6bufg[INET6_ADDRSTRLEN];
1574 			char ip6bufm[INET6_ADDRSTRLEN];
1575 			struct sockaddr_in6 *sin6;
1576 
1577 			sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
1578 			nd6log((LOG_ERR, "nd6_prefix_onlink: failed to add "
1579 			    "route for a prefix (%s/%d) on %s, gw=%s, mask=%s, "
1580 			    "flags=%lx errno = %d\n",
1581 			    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1582 			    pr->ndpr_plen, if_name(pr->ndpr_ifp),
1583 			    ip6_sprintf(ip6bufg, &sin6->sin6_addr),
1584 			    ip6_sprintf(ip6bufm, &mask6.sin6_addr),
1585 			    rtflags, error));
1586 
1587 			/* Save last error to return, see rtinit(). */
1588 			a_failure = error;
1589 		}
1590 
1591 		if (rt != NULL) {
1592 			RT_LOCK(rt);
1593 			RT_REMREF(rt);
1594 			RT_UNLOCK(rt);
1595 		}
1596 	}
1597 
1598 	/* Return the last error we got. */
1599 	return (a_failure);
1600 }
1601 
1602 static int
1603 nd6_prefix_onlink(struct nd_prefix *pr)
1604 {
1605 	struct ifaddr *ifa;
1606 	struct ifnet *ifp = pr->ndpr_ifp;
1607 	struct nd_prefix *opr;
1608 	int error = 0;
1609 	char ip6buf[INET6_ADDRSTRLEN];
1610 
1611 	/* sanity check */
1612 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
1613 		nd6log((LOG_ERR,
1614 		    "nd6_prefix_onlink: %s/%d is already on-link\n",
1615 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1616 		    pr->ndpr_plen));
1617 		return (EEXIST);
1618 	}
1619 
1620 	/*
1621 	 * Add the interface route associated with the prefix.  Before
1622 	 * installing the route, check if there's the same prefix on another
1623 	 * interface, and the prefix has already installed the interface route.
1624 	 * Although such a configuration is expected to be rare, we explicitly
1625 	 * allow it.
1626 	 */
1627 	LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) {
1628 		if (opr == pr)
1629 			continue;
1630 
1631 		if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
1632 			continue;
1633 
1634 		if (opr->ndpr_plen == pr->ndpr_plen &&
1635 		    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1636 		    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen))
1637 			return (0);
1638 	}
1639 
1640 	/*
1641 	 * We prefer link-local addresses as the associated interface address.
1642 	 */
1643 	/* search for a link-local addr */
1644 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
1645 	    IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
1646 	if (ifa == NULL) {
1647 		/* XXX: freebsd does not have ifa_ifwithaf */
1648 		IF_ADDR_RLOCK(ifp);
1649 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1650 			if (ifa->ifa_addr->sa_family == AF_INET6)
1651 				break;
1652 		}
1653 		if (ifa != NULL)
1654 			ifa_ref(ifa);
1655 		IF_ADDR_RUNLOCK(ifp);
1656 		/* should we care about ia6_flags? */
1657 	}
1658 	if (ifa == NULL) {
1659 		/*
1660 		 * This can still happen, when, for example, we receive an RA
1661 		 * containing a prefix with the L bit set and the A bit clear,
1662 		 * after removing all IPv6 addresses on the receiving
1663 		 * interface.  This should, of course, be rare though.
1664 		 */
1665 		nd6log((LOG_NOTICE,
1666 		    "nd6_prefix_onlink: failed to find any ifaddr"
1667 		    " to add route for a prefix(%s/%d) on %s\n",
1668 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1669 		    pr->ndpr_plen, if_name(ifp)));
1670 		return (0);
1671 	}
1672 
1673 	error = nd6_prefix_onlink_rtrequest(pr, ifa);
1674 
1675 	if (ifa != NULL)
1676 		ifa_free(ifa);
1677 
1678 	return (error);
1679 }
1680 
1681 static int
1682 nd6_prefix_offlink(struct nd_prefix *pr)
1683 {
1684 	int error = 0;
1685 	struct ifnet *ifp = pr->ndpr_ifp;
1686 	struct nd_prefix *opr;
1687 	struct sockaddr_in6 sa6, mask6;
1688 	struct rtentry *rt;
1689 	char ip6buf[INET6_ADDRSTRLEN];
1690 	int fibnum, a_failure;
1691 
1692 	/* sanity check */
1693 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
1694 		nd6log((LOG_ERR,
1695 		    "nd6_prefix_offlink: %s/%d is already off-link\n",
1696 		    ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
1697 		    pr->ndpr_plen));
1698 		return (EEXIST);
1699 	}
1700 
1701 	bzero(&sa6, sizeof(sa6));
1702 	sa6.sin6_family = AF_INET6;
1703 	sa6.sin6_len = sizeof(sa6);
1704 	bcopy(&pr->ndpr_prefix.sin6_addr, &sa6.sin6_addr,
1705 	    sizeof(struct in6_addr));
1706 	bzero(&mask6, sizeof(mask6));
1707 	mask6.sin6_family = AF_INET6;
1708 	mask6.sin6_len = sizeof(sa6);
1709 	bcopy(&pr->ndpr_mask, &mask6.sin6_addr, sizeof(struct in6_addr));
1710 
1711 	a_failure = 0;
1712 	for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
1713 		rt = NULL;
1714 		error = in6_rtrequest(RTM_DELETE, (struct sockaddr *)&sa6, NULL,
1715 		    (struct sockaddr *)&mask6, 0, &rt, fibnum);
1716 		if (error == 0) {
1717 			/* report the route deletion to the routing socket. */
1718 			if (rt != NULL)
1719 				nd6_rtmsg(RTM_DELETE, rt);
1720 		} else {
1721 			/* Save last error to return, see rtinit(). */
1722 			a_failure = error;
1723 		}
1724 		if (rt != NULL) {
1725 			RTFREE(rt);
1726 		}
1727 	}
1728 	error = a_failure;
1729 	a_failure = 1;
1730 	if (error == 0) {
1731 		pr->ndpr_stateflags &= ~NDPRF_ONLINK;
1732 
1733 		/*
1734 		 * There might be the same prefix on another interface,
1735 		 * the prefix which could not be on-link just because we have
1736 		 * the interface route (see comments in nd6_prefix_onlink).
1737 		 * If there's one, try to make the prefix on-link on the
1738 		 * interface.
1739 		 */
1740 		LIST_FOREACH(opr, &V_nd_prefix, ndpr_entry) {
1741 			if (opr == pr)
1742 				continue;
1743 
1744 			if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0)
1745 				continue;
1746 
1747 			/*
1748 			 * KAME specific: detached prefixes should not be
1749 			 * on-link.
1750 			 */
1751 			if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0)
1752 				continue;
1753 
1754 			if (opr->ndpr_plen == pr->ndpr_plen &&
1755 			    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
1756 			    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
1757 				int e;
1758 
1759 				if ((e = nd6_prefix_onlink(opr)) != 0) {
1760 					nd6log((LOG_ERR,
1761 					    "nd6_prefix_offlink: failed to "
1762 					    "recover a prefix %s/%d from %s "
1763 					    "to %s (errno = %d)\n",
1764 					    ip6_sprintf(ip6buf,
1765 						&opr->ndpr_prefix.sin6_addr),
1766 					    opr->ndpr_plen, if_name(ifp),
1767 					    if_name(opr->ndpr_ifp), e));
1768 				} else
1769 					a_failure = 0;
1770 			}
1771 		}
1772 	} else {
1773 		/* XXX: can we still set the NDPRF_ONLINK flag? */
1774 		nd6log((LOG_ERR,
1775 		    "nd6_prefix_offlink: failed to delete route: "
1776 		    "%s/%d on %s (errno = %d)\n",
1777 		    ip6_sprintf(ip6buf, &sa6.sin6_addr), pr->ndpr_plen,
1778 		    if_name(ifp), error));
1779 	}
1780 
1781 	if (a_failure)
1782 		lltable_prefix_free(AF_INET6, (struct sockaddr *)&sa6,
1783 		    (struct sockaddr *)&mask6, LLE_STATIC);
1784 
1785 	return (error);
1786 }
1787 
1788 static struct in6_ifaddr *
1789 in6_ifadd(struct nd_prefixctl *pr, int mcast)
1790 {
1791 	struct ifnet *ifp = pr->ndpr_ifp;
1792 	struct ifaddr *ifa;
1793 	struct in6_aliasreq ifra;
1794 	struct in6_ifaddr *ia, *ib;
1795 	int error, plen0;
1796 	struct in6_addr mask;
1797 	int prefixlen = pr->ndpr_plen;
1798 	int updateflags;
1799 	char ip6buf[INET6_ADDRSTRLEN];
1800 
1801 	in6_prefixlen2mask(&mask, prefixlen);
1802 
1803 	/*
1804 	 * find a link-local address (will be interface ID).
1805 	 * Is it really mandatory? Theoretically, a global or a site-local
1806 	 * address can be configured without a link-local address, if we
1807 	 * have a unique interface identifier...
1808 	 *
1809 	 * it is not mandatory to have a link-local address, we can generate
1810 	 * interface identifier on the fly.  we do this because:
1811 	 * (1) it should be the easiest way to find interface identifier.
1812 	 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
1813 	 * for multiple addresses on a single interface, and possible shortcut
1814 	 * of DAD.  we omitted DAD for this reason in the past.
1815 	 * (3) a user can prevent autoconfiguration of global address
1816 	 * by removing link-local address by hand (this is partly because we
1817 	 * don't have other way to control the use of IPv6 on an interface.
1818 	 * this has been our design choice - cf. NRL's "ifconfig auto").
1819 	 * (4) it is easier to manage when an interface has addresses
1820 	 * with the same interface identifier, than to have multiple addresses
1821 	 * with different interface identifiers.
1822 	 */
1823 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
1824 	if (ifa)
1825 		ib = (struct in6_ifaddr *)ifa;
1826 	else
1827 		return NULL;
1828 
1829 	/* prefixlen + ifidlen must be equal to 128 */
1830 	plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
1831 	if (prefixlen != plen0) {
1832 		ifa_free(ifa);
1833 		nd6log((LOG_INFO, "in6_ifadd: wrong prefixlen for %s "
1834 		    "(prefix=%d ifid=%d)\n",
1835 		    if_name(ifp), prefixlen, 128 - plen0));
1836 		return NULL;
1837 	}
1838 
1839 	/* make ifaddr */
1840 	in6_prepare_ifra(&ifra, &pr->ndpr_prefix.sin6_addr, &mask);
1841 
1842 	IN6_MASK_ADDR(&ifra.ifra_addr.sin6_addr, &mask);
1843 	/* interface ID */
1844 	ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
1845 	    (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
1846 	ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
1847 	    (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
1848 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1849 	    (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
1850 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1851 	    (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
1852 	ifa_free(ifa);
1853 
1854 	/* lifetimes. */
1855 	ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime;
1856 	ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime;
1857 
1858 	/* XXX: scope zone ID? */
1859 
1860 	ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
1861 
1862 	/*
1863 	 * Make sure that we do not have this address already.  This should
1864 	 * usually not happen, but we can still see this case, e.g., if we
1865 	 * have manually configured the exact address to be configured.
1866 	 */
1867 	ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
1868 	    &ifra.ifra_addr.sin6_addr);
1869 	if (ifa != NULL) {
1870 		ifa_free(ifa);
1871 		/* this should be rare enough to make an explicit log */
1872 		log(LOG_INFO, "in6_ifadd: %s is already configured\n",
1873 		    ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr));
1874 		return (NULL);
1875 	}
1876 
1877 	/*
1878 	 * Allocate ifaddr structure, link into chain, etc.
1879 	 * If we are going to create a new address upon receiving a multicasted
1880 	 * RA, we need to impose a random delay before starting DAD.
1881 	 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
1882 	 */
1883 	updateflags = 0;
1884 	if (mcast)
1885 		updateflags |= IN6_IFAUPDATE_DADDELAY;
1886 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
1887 		nd6log((LOG_ERR,
1888 		    "in6_ifadd: failed to make ifaddr %s on %s (errno=%d)\n",
1889 		    ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr),
1890 		    if_name(ifp), error));
1891 		return (NULL);	/* ifaddr must not have been allocated. */
1892 	}
1893 
1894 	ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
1895 	/*
1896 	 * XXXRW: Assumption of non-NULLness here might not be true with
1897 	 * fine-grained locking -- should we validate it?  Or just return
1898 	 * earlier ifa rather than looking it up again?
1899 	 */
1900 	return (ia);		/* this is always non-NULL  and referenced. */
1901 }
1902 
1903 /*
1904  * ia0 - corresponding public address
1905  */
1906 int
1907 in6_tmpifadd(const struct in6_ifaddr *ia0, int forcegen, int delay)
1908 {
1909 	struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
1910 	struct in6_ifaddr *newia;
1911 	struct in6_aliasreq ifra;
1912 	int error;
1913 	int trylimit = 3;	/* XXX: adhoc value */
1914 	int updateflags;
1915 	u_int32_t randid[2];
1916 	time_t vltime0, pltime0;
1917 
1918 	in6_prepare_ifra(&ifra, &ia0->ia_addr.sin6_addr,
1919 	    &ia0->ia_prefixmask.sin6_addr);
1920 
1921 	ifra.ifra_addr = ia0->ia_addr;	/* XXX: do we need this ? */
1922 	/* clear the old IFID */
1923 	IN6_MASK_ADDR(&ifra.ifra_addr.sin6_addr,
1924 	    &ifra.ifra_prefixmask.sin6_addr);
1925 
1926   again:
1927 	if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
1928 	    (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
1929 		nd6log((LOG_NOTICE, "in6_tmpifadd: failed to find a good "
1930 		    "random IFID\n"));
1931 		return (EINVAL);
1932 	}
1933 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
1934 	    (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
1935 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
1936 	    (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
1937 
1938 	/*
1939 	 * in6_get_tmpifid() quite likely provided a unique interface ID.
1940 	 * However, we may still have a chance to see collision, because
1941 	 * there may be a time lag between generation of the ID and generation
1942 	 * of the address.  So, we'll do one more sanity check.
1943 	 */
1944 
1945 	if (in6_localip(&ifra.ifra_addr.sin6_addr) != 0) {
1946 		if (trylimit-- > 0) {
1947 			forcegen = 1;
1948 			goto again;
1949 		}
1950 
1951 		/* Give up.  Something strange should have happened.  */
1952 		nd6log((LOG_NOTICE, "in6_tmpifadd: failed to "
1953 		    "find a unique random IFID\n"));
1954 		return (EEXIST);
1955 	}
1956 
1957 	/*
1958 	 * The Valid Lifetime is the lower of the Valid Lifetime of the
1959          * public address or TEMP_VALID_LIFETIME.
1960 	 * The Preferred Lifetime is the lower of the Preferred Lifetime
1961          * of the public address or TEMP_PREFERRED_LIFETIME -
1962          * DESYNC_FACTOR.
1963 	 */
1964 	if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1965 		vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
1966 		    (ia0->ia6_lifetime.ia6t_vltime -
1967 		    (time_uptime - ia0->ia6_updatetime));
1968 		if (vltime0 > V_ip6_temp_valid_lifetime)
1969 			vltime0 = V_ip6_temp_valid_lifetime;
1970 	} else
1971 		vltime0 = V_ip6_temp_valid_lifetime;
1972 	if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1973 		pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
1974 		    (ia0->ia6_lifetime.ia6t_pltime -
1975 		    (time_uptime - ia0->ia6_updatetime));
1976 		if (pltime0 > V_ip6_temp_preferred_lifetime - V_ip6_desync_factor){
1977 			pltime0 = V_ip6_temp_preferred_lifetime -
1978 			    V_ip6_desync_factor;
1979 		}
1980 	} else
1981 		pltime0 = V_ip6_temp_preferred_lifetime - V_ip6_desync_factor;
1982 	ifra.ifra_lifetime.ia6t_vltime = vltime0;
1983 	ifra.ifra_lifetime.ia6t_pltime = pltime0;
1984 
1985 	/*
1986 	 * A temporary address is created only if this calculated Preferred
1987 	 * Lifetime is greater than REGEN_ADVANCE time units.
1988 	 */
1989 	if (ifra.ifra_lifetime.ia6t_pltime <= V_ip6_temp_regen_advance)
1990 		return (0);
1991 
1992 	/* XXX: scope zone ID? */
1993 
1994 	ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
1995 
1996 	/* allocate ifaddr structure, link into chain, etc. */
1997 	updateflags = 0;
1998 	if (delay)
1999 		updateflags |= IN6_IFAUPDATE_DADDELAY;
2000 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
2001 		return (error);
2002 
2003 	newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
2004 	if (newia == NULL) {	/* XXX: can it happen? */
2005 		nd6log((LOG_ERR,
2006 		    "in6_tmpifadd: ifa update succeeded, but we got "
2007 		    "no ifaddr\n"));
2008 		return (EINVAL); /* XXX */
2009 	}
2010 	newia->ia6_ndpr = ia0->ia6_ndpr;
2011 	newia->ia6_ndpr->ndpr_refcnt++;
2012 	ifa_free(&newia->ia_ifa);
2013 
2014 	/*
2015 	 * A newly added address might affect the status of other addresses.
2016 	 * XXX: when the temporary address is generated with a new public
2017 	 * address, the onlink check is redundant.  However, it would be safe
2018 	 * to do the check explicitly everywhere a new address is generated,
2019 	 * and, in fact, we surely need the check when we create a new
2020 	 * temporary address due to deprecation of an old temporary address.
2021 	 */
2022 	pfxlist_onlink_check();
2023 
2024 	return (0);
2025 }
2026 
2027 static int
2028 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
2029 {
2030 	if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
2031 		ndpr->ndpr_preferred = 0;
2032 	else
2033 		ndpr->ndpr_preferred = time_uptime + ndpr->ndpr_pltime;
2034 	if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
2035 		ndpr->ndpr_expire = 0;
2036 	else
2037 		ndpr->ndpr_expire = time_uptime + ndpr->ndpr_vltime;
2038 
2039 	return 0;
2040 }
2041 
2042 static void
2043 in6_init_address_ltimes(struct nd_prefix *new, struct in6_addrlifetime *lt6)
2044 {
2045 	/* init ia6t_expire */
2046 	if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
2047 		lt6->ia6t_expire = 0;
2048 	else {
2049 		lt6->ia6t_expire = time_uptime;
2050 		lt6->ia6t_expire += lt6->ia6t_vltime;
2051 	}
2052 
2053 	/* init ia6t_preferred */
2054 	if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
2055 		lt6->ia6t_preferred = 0;
2056 	else {
2057 		lt6->ia6t_preferred = time_uptime;
2058 		lt6->ia6t_preferred += lt6->ia6t_pltime;
2059 	}
2060 }
2061 
2062 /*
2063  * Delete all the routing table entries that use the specified gateway.
2064  * XXX: this function causes search through all entries of routing table, so
2065  * it shouldn't be called when acting as a router.
2066  */
2067 void
2068 rt6_flush(struct in6_addr *gateway, struct ifnet *ifp)
2069 {
2070 
2071 	/* We'll care only link-local addresses */
2072 	if (!IN6_IS_ADDR_LINKLOCAL(gateway))
2073 		return;
2074 
2075 	/* XXX Do we really need to walk any but the default FIB? */
2076 	rt_foreach_fib_walk(AF_INET6, NULL, rt6_deleteroute, (void *)gateway);
2077 }
2078 
2079 static int
2080 rt6_deleteroute(struct rtentry *rt, void *arg)
2081 {
2082 #define SIN6(s)	((struct sockaddr_in6 *)s)
2083 	struct in6_addr *gate = (struct in6_addr *)arg;
2084 
2085 	if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
2086 		return (0);
2087 
2088 	if (!IN6_ARE_ADDR_EQUAL(gate, &SIN6(rt->rt_gateway)->sin6_addr)) {
2089 		return (0);
2090 	}
2091 
2092 	/*
2093 	 * Do not delete a static route.
2094 	 * XXX: this seems to be a bit ad-hoc. Should we consider the
2095 	 * 'cloned' bit instead?
2096 	 */
2097 	if ((rt->rt_flags & RTF_STATIC) != 0)
2098 		return (0);
2099 
2100 	/*
2101 	 * We delete only host route. This means, in particular, we don't
2102 	 * delete default route.
2103 	 */
2104 	if ((rt->rt_flags & RTF_HOST) == 0)
2105 		return (0);
2106 
2107 	return (in6_rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
2108 	    rt_mask(rt), rt->rt_flags | RTF_RNH_LOCKED, NULL, rt->rt_fibnum));
2109 #undef SIN6
2110 }
2111 
2112 int
2113 nd6_setdefaultiface(int ifindex)
2114 {
2115 	int error = 0;
2116 
2117 	if (ifindex < 0 || V_if_index < ifindex)
2118 		return (EINVAL);
2119 	if (ifindex != 0 && !ifnet_byindex(ifindex))
2120 		return (EINVAL);
2121 
2122 	if (V_nd6_defifindex != ifindex) {
2123 		V_nd6_defifindex = ifindex;
2124 		if (V_nd6_defifindex > 0)
2125 			V_nd6_defifp = ifnet_byindex(V_nd6_defifindex);
2126 		else
2127 			V_nd6_defifp = NULL;
2128 
2129 		/*
2130 		 * Our current implementation assumes one-to-one maping between
2131 		 * interfaces and links, so it would be natural to use the
2132 		 * default interface as the default link.
2133 		 */
2134 		scope6_setdefault(V_nd6_defifp);
2135 	}
2136 
2137 	return (error);
2138 }
2139